Platform

The Blue Mesh platform

AI Agents, Voice Agents, Enterprise RAG and model training, built on one canvas and run in your private cloud, on premises or fully air-gapped, with role-based access and an audit trail on every decision. Integrates with 300+ existing systems and tools.

01
TRIGGER
input.received
webhook
02
AGENT
plan.compose
reasoning
03
TOOL
web.search
live data
04
LOOP
iterate.until(done)
×2
05
EGRESS
response.sent
traced

A workflow's first run

The real interface running a newly saved workflow: one request in, the answer back with its latency, every step green (1.25x speed).

THE PROBLEM

Enterprise AI stalls where the data cannot go

// Data that cannot leave

The workflows worth automating run on customer records, contracts and calls. Public AI services need that data to leave your boundary, and the security review that follows is often where the project stops.

// One tool per job

A chatbot here, a transcription service there, a separate search tool: each with its own access rules, logs and bill, and none of them able to finish the work behind the request.

// No record of what the AI did

When an AI step goes wrong, there is no trace of what it read, what it decided and who let it act, which is the first thing an auditor asks for.

THE SOLUTION

One platform, in your environment, for the whole job

Blue Mesh puts agents, voice agents, retrieval and model training on one canvas that runs where your data already lives. Each capability below is part of the same platform, under the same access rules and the same audit trail.

ORCHESTRATE

What an agent is built from

A Leader Agent decomposes each task and directs specialist agents end to end, with no manual handoffs. Approvals can be placed at every level, including inside nested workflows, so a consequential step pauses for a person while everything under it runs.

Voice agents

Voice agents on real phone calls over SIP, on your own models, escalating to a person when needed.

300+ Systems and Tools

Pre-built integrations with the systems you already run, so agents act through them under the credentials your team already governs.

Smart Web Search

Web search as a workflow step, pulling live information into an agent's context on demand. A step your security team can see, and switch off in an air-gapped deployment.

Loop Node

A step that repeats until a condition you define is met, with the iteration count in the trace, and no code required.

TRAIN

Train on your data, answer from your documents

BM Oasis

Fine-tune models on your own data, inside your own environment. The weights are yours.

Enterprise RAG

Retrieval-augmented answers from your indexed data, with the source named in every reply.

CONNECT

Connect the systems you already run

OpenAI-API Compatibility

A BM Studio deployment answers as an OpenAI-compatible API, so existing applications point at it unchanged.

Webhooks

Incoming webhook endpoints let your systems start a workflow on a real event: an order, a ticket, an alert.

Custom API Tools

Your own internal APIs plug in as reusable tools that agents call, under credentials your team governs.

OpenRouter

Optional connectivity to OpenRouter for hosted models through one interface, enabled by you, off when air-gapped.

OPERATE

Every step visible, every token counted

Explainable AI

Each agent step shown as it runs: the reasoning, the tool called, the result. The audit trail keeps the same record.

Token Tracking

Token consumption recorded per request, so the cost of a workflow is known per run, not discovered at month end.

DEPLOY AND GOVERN

Where it runs, and who is in control

There is no metered public API and no self-serve tier. Blue Mesh deploys into your environment, so a first deployment starts with a conversation about that environment.

  • 01Deploys in your private cloud, your data centre, or fully air-gapped; model weights and inference stay inside your network boundary.
  • 02Role-based access decides who can build, run, approve and read.
  • 03Every decision and result is logged for compliance audit: what ran, what it read, what it did, and who approved it.
  • 04Usage is yours to control: the organisation decides who in it can use what, and how much.
THE RESULT

Already in production

Vaspian · US telecom provider

950K

transcript requests each month

Read the case study →

Spectrotel · US managed network provider

500K

network alerts triaged each month

Read the case study →
Where to go next
  • BM Studio The canvas these nodes, tools and loops are assembled on, drag and drop, with no code.
  • BM Architect Describe a workflow in plain language and have it designed, tested and deployed, with a demo to show for it.
  • BM Oasis Fine-tuning on your own data, and a straight answer on who keeps the weights that come out of it.
  • BM in a Box Agents and RAG as a pre-configured appliance on NVIDIA DGX hardware, running with no outbound connection.

Answers only from the pages on this site. Nothing you type here is stored.

See the platform run on a workflow of yours

Bring a real workflow. In one session we will build an agent for it, live, and show you exactly where your team stays in control.